chapter nine

9 More stable value-based methods

 

In this chapter:

  • You improve on the methods you learned in the previous chapter by making them more stable and therefore less prone to divergence.
  • You explore advanced value-based deep reinforcement learning methods, and the many components that make value-based methods better.
  • You solve the cart-pole environment in a fewer number of samples, and with more reliable and consistent results.

Let thy step be slow and steady, that thou stumble not.

— Tokugawa Ieyasu, Founder and first shōgun of the Tokugawa shogunate of Japan and one of the three unifiers of Japan.

9.1     DQN: Making reinforcement learning more like supervised learning

9.1.1     Common problems in value-based deep reinforcement learning

9.1.2     Using target networks

9.1.3     Using larger networks

9.1.4     Using experience replay

9.1.5     Using other exploration strategies

9.2     Double DQN: Mitigating the overestimation of action-value functions

9.2.1     The problem of overestimation, take two

9.2.2     Separating action selection and action evaluation

9.2.3     A solution

9.2.4     A more practical solution